Re-hospitalization risk predicting method based on deep learning hybrid model

Abstract

The invention discloses a re-hospitalization risk predicting method based on a deep learning hybrid model. The method comprises the following steps of 1, collecting a data set which comprises a patient personal characteristic and an outer environment characteristic; 2, performing characteristic grouping and preprocessing, dividing the characteristics to static characteristics and time sequence characteristics; 3, mining the time sequence characteristics, performing statistics analysis on the time sequence characteristics, and establishing an LDA model and a bidirectional LSTM model; 4, performing characteristic splicing, fusing the static characteristics and the time sequence characteristics after characteristic engineering processing, and using the fused characteristic as input of a CNN model; and 5, establishing the CNN model for predicting the patient re-hospitalization risk. The method is based on a deep learning algorithm. Researching and analysis are performed on patient health medical big data and the outer environment, thereby establishing the re-hospitalization risk predicting model, facilitating reasonable medical resource arrangement by a medical organization, supplyingbetter medical service to the patient, and improving efficiency and accuracy in performing re-hospitalization risk identification of a participant by an insurance organization.

Description

technical field [0001] The present invention relates to the medical and health field and machine learning technology, in particular to a rehospitalization risk prediction method based on LDA, LSTM, and CNN hybrid models. Background technique [0002] With the continuous improvement of medical informatization, my country's medical institutions have entered the era of informatization and digitalization. The medical and health field has accumulated a large amount of data, providing a solid data foundation for "artificial intelligence + medical" research. At the same time, precision health care has been valued by more and more countries and institutions, and readmission risk prediction is an important research direction of precision health care. [0003] Accurate and personalized readmission risk prediction has great application value for improving the level of medical public services and medical insurance planning. In terms of medical public services, rehospitalization risk pr...

AI Technical Summary

Problems solved by technology

[0004] Analyzing the characteristics of health care big data and the research status of the rehospitalization risk prediction model, there are still some problems to be studied in the current work, which are mainly reflected in the following three aspects: (1) The rehospitalization risk prediction model mainly considers the individual characteristics of patients and ignores The impact of the external environment on patient rehospitalization; (2) Insufficient information mining on patient disease change trends, treatment paths, and disease similarities; (3) At present, traditional machine learning algorithms are mainly used to build rehospitalization risk prediction models, while Traditional machine learning algorithms based on tree models are extremely inefficient when dealing with large sample sizes

Images